Track Operational Metrics and Generate Insights

This skill designs comprehensive operational metrics tracking systems. SMART metrics design principles ensure each metric is specific, measurable, achievable, relevant, and time-bound, organized into a three-tier hierarchy: strategic (revenue growth, market share, LTV), tactical (conversion rate, CAC, employee productivity), and operational (response times, error rates, resource utilization).

Essential metric categories are implemented with concrete formulas. Financial metrics compute MRR (sum of active subscription monthly values), ARR growth rate, and LTV:CAC ratio. Operational efficiency uses a SQL query computing average and p95 response time plus error rate (percentage of 5xx responses) per day over a 30-day window. Customer experience metrics compute NPS (promoters minus detractors as a percentage of total respondents), churn rate accounting for new customer additions, and CAC as combined sales/marketing spend divided by new customers.

def calculate_ltv_cac_ratio(self, ltv, cac):
    return ltv / cac if cac > 0 else 0

calculateNPS: (responses) => {
    const promoters = responses.filter(r => r.score >= 9).length;
    const detractors = responses.filter(r => r.score <= 6).length;
    return ((promoters - detractors) / responses.length) * 100;
}

Collection architecture uses an async MetricsCollector with configurable intervals (real-time, hourly, daily) that gathers system metrics (CPU, memory, connections, queue depth) and separately calculates business KPIs (daily revenue, transaction count, average transaction value, unique customers) from a trailing 24-hour window of raw transaction data. Dashboard design follows an executive layout with sections for revenue overview (metric cards with period comparison, trend charts) and operational health (a threshold-colored gauge for uptime, a status grid for API response time/error rate/queue health). Alerting is configured declaratively with per-metric condition, timeframe, severity, and notification channel (e.g. revenue drop over 10% in 24h routed to email/Slack, error rate over 5% in 5 minutes routed to PagerDuty/Slack as critical). Best practices cover data quality/governance (validation at collection, documented calculation methods, lineage documentation, versioned metric definitions), performance optimization (pre-aggregation, time-series indexing, retention policies, partitioning), and stakeholder engagement (audience-tailored dashboards, benchmarked context, regular review sessions). Advanced analytics covers a linear-regression-based revenue forecasting class predicting future values with a confidence score and trend direction.

Use this skill when designing or extending an operational metrics tracking system - defining a SMART metric hierarchy, implementing financial/efficiency/customer-experience metric calculations, building a real-time collection pipeline, designing executive dashboards, configuring threshold-based alerting, or forecasting metric trends.

It is not the right fit for one-off ad-hoc reporting with no ongoing tracking need, or for organizations without the underlying data infrastructure (transaction logs, API logs, subscription records) to compute these metrics reliably.

Input: the business's strategic objectives and the raw operational data available (transactions, API logs, subscriptions, customer surveys). Output: a SMART metric hierarchy across strategic/tactical/operational tiers, implemented calculation logic for key metrics (MRR, NPS, churn, error rate), a real-time collection pipeline, an executive dashboard configuration, threshold-based alert rules, and optionally a linear-regression forecast of metric trends.

Metrics calculations are implemented in Python (asyncio for the collection pipeline, scikit-learn for forecasting) and SQL for time-series aggregation, with dashboard and alert configuration expressed in JSON/YAML for integration with monitoring and notification systems (email, Slack, PagerDuty).

Operations and business intelligence teams building metrics tracking systems - particularly those needing a structured metric hierarchy, real-time collection pipelines, and threshold-based alerting tied to business and operational health.